Hacker Newsnew | past | comments | ask | show | jobs | submit | MattRix's commentslogin

I don't see any reason all the things you mentioned cannot be taught to an LLM using reinforcement learning. The newer models are already much better at this stuff, I see no reason that trajectory will change.

Hence I wrote: "Structured data of actual usage would be needed for AI to ingest, in a way that it can derive intent from every action and identify friction."

The challenge is more about providing actual, objective data to learn this.

The LLM will not be able to decide what is "good UX" and "bad UX", it will have to ingest this information in vast detail from judgement provided by humans, which is also varying alot.

Case in point: Coming from that field, I personally don't have a good user experience using iOS, a majority of end-users will strongly disagree.

An LLM would need to be trained to handle both cases to judge the same human interface, so it would need alot of data to understand both views...


The human UI designer has the benefit of being and to run experiments on themselves, but only n=1. for any problem where that is insufficient or even misleading that benefit disappears. I’d argue that UI is an area where you need n>>1.

That's what UX user studies are for, which is a process of humans observing other humans using the product and taking notes. It's far more common than people seem to believe.

(also: UX is not UI, but I guess this goes without saying...)


I don't think this is true because Opus 5 was bad and Opus 5.5 was really good, and presumably those both used the same pre-train? So instead it's more of a post-training thing?

If I don't give him my money, then that less money he can put towards bad causes. It gives him less of a platform to espouse his views too. There are plenty of reasons to not give bad people money, and most of them are valid.

On top of that, if his profits going down causes him to genuinely reconsider his moral principles, that isn't inherently unprincipled! He may just have not thought about it seriously until now.

But either way, that's fine. I would much rather give money to someone who makes decisions partly for financial/selfish reasons (like most of us do!) than someone who is openly bigoted or transphobic.


The fact that it's expensive has nothing to do with how useful it is though... I know plenty of people who max out their $200 plans every week, and they are working on stuff a lot less financially valuable than what Apple employees work on.

But does the spending on the ai result in more revenue for the company?

I know people who spend that on lottery tickets. Obviously rational.

I don't think you actually know anyone spending $10,000/yr on lotto tickets.

But it's telling that you claim to while treating this as some sort of equivalence.


You caught me. I would guess though, that at the end user/retail level, more money is spent on lottery tickets than AI subscriptions. Does that work better?

You are misunderstanding the quote.

In which way?

It was not about any of the things you stated. It was about the simple fact that DRAM in data centers get used more than personal/local DRAM.

Yes. And pretending that other parts I mentioned do not matter is exactly lying by omission

They don't matter though... there is no lying by omission, what?


I'm not sure what you think happened, but there was no lying by omission because the discussion was not about pros and cons of local models or anything of the sort.

It has nothing to do with his POV or even ego. It was an objectively true fact. GPUs in data centers do get more utilization than those used locally. That's all he was saying.

And engines in busses objectively get more utilization than the combustion engine in your vehicle.

Also true

Something can be a true fact and still be an ill-advised or tone deaf thing to say.

Especially when you take into account that people are largely unconcerned about true facts, and will gladly misquote or misinterpret you just in order to be outraged.

It wasn't ill advised or tone deaf, it was taken by the media out of context. You can't take stuff completely out of context and then complain that it's tone deaf.

This is techno utilitarianism.

Also, a lot of things are true but what truths you choose to broadcast can tell us what you care about. E.g it's also true that we'd all be richer if sam altman shared is wealth with us, lets take a few seconds to think why he's not saying this truth, after all it's an objectively true fact.


You can think whatever you want about the guy, but when you willfully misinterpret his words because it fits with your view of what he secretly believes anyway, I think that's still a bad thing for you to do.

Is it really "all he was saying" when he's single handedly responsible for creating a worldwide RAM and NAND shortage?

He said/implied that we would have shortage either way (without data center buildout, but everyone who wants access to LLM does private/local inference), which doesn't sound far fetched.

Single handedly? I'd say the the people using (or forcing others to use) the models, therefore justifying the need for them, play a part in it.


Sure, he wasn't implying anything. He just recited this fact completely out of context and wasted airtime.

No, he was not saying that was his goal! That's what everyone seems to misunderstand about this quote.

It's like a highrise developer saying "people think our buildings take up a lot of land, but actually houses are much worse from a land use perspective". Which is objectively true, and yet does not mean the developer thinks the entire world should be turned into houses, or even that they dislike houses!


Except he was the developer who signed a deal with land owners to buy 40% of the land with the intention of building houses to rent and caused condo prices to 5x.

No, it's more like he was just clearing up the misconception that "cities would be better places to live without all the highrises in them".

As self driving cars start to take over, this will become a non-issue, since they can drive more cautiously and don't have blind spots.


Meanwhile, can we please have laws that require trucks to have an electronic assistant checking if there's someone next to the vehicle when talking a turn?


Because they will start getting sued or arrested for harms the models they host cause. Anyone who thinks this won't happen isn't paying attention. And if you think "we'll just torrent them", let's just say the powers that be are gonna come down harder on this stuff than anything in the past. This will get taken a lot more seriously than torrenting a movie.


That'd be like successfully suing a gas station because your self-driving car filled up there and then killed someone.

It is the driver/user of the car who is liable, and possibly the manufacturer if it's defective.


hah didn't think my game would be listed there, but it was! https://fivefold.ca


Nice. I assume you've seen the Sudoku puzzles at Logic Masters Germany? Heavily reminiscent.

One suggestion, and maybe I'm just holding it wrong: when I started the daily puzzle, I had to scroll down to read through the rules, which I assume starts the timer while I'm still learning the symbols. It would be nice to have an interim page with a summary of the symbols so the timer doesn't start immediately.


Hah I completely agree! I actually already had that on my task list, I'm adding that exact feature to both the webapp and mobile version this weekend.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: